Textural analysis of ocean surface images

The bases of a radar image analysis strategy are presented in the context of wave or swell phenomena description from remote sensing ocean images. First of all, local characterization is a necessity especially for coastal areas where various non-stationarities are revealed. The authors chose to apply to these types of images a classical detection-estimation scheme in order to provide focus areas and then relevant features of the surface. Some spatial and 2D frequency processing tools are proposed to analyse the local texture. They give various measures such as anisotropy, contrast, dominant orientation, dominant wave length, length of crest, frequency decomposition, global variation of the direction and frequency. The importance of the a priori knowledge needed in most of the process is highlighted and can be linked in some extent to a physical interpretation